00:15 <smaeul> apritzel: good question. it could be put in either the ccu driver directly, or in the rtc driver with of_clk_get_from_provider()

00:17 <smaeul> I'm not sure how we decide which parent to use for rtc-32k

00:20 <bauen1> jakllsch: it's a board from https://letstrust.de/

00:20 <smaeul> what is the problem you are seeing on H616? are there really no other osc32k consumers and iosc gets gated?

00:21 <bauen1> anarsoul: that's interesting, but i would have expected to at least be able to issue commands, however it seems that i can't even read the registers over SPI correctly, and i'm not even talking about requesting a locality

00:22 <anarsoul> yeah, maybe you fried it :)

00:23 <bauen1> anarsoul: now my breadboard setup with some jumpers probably isn't helping, but at this time i'm still suspecting a timing error in the spi, as the chip seems to respond with roughly the right structure, but is most times returning very wrong values (as in 0xff 0xff 0xff 0xff for a register that must exist)

00:23 <bauen1> anarsoul: yes, and i did talk to it with a broken sunxi-spi driver so it wouldn't even surprise me if that caused some 5v levels to reacht the poor chip, it might be time to dump another 25€ for another one

00:25 <bauen1> or perhaps write a few mails to infineon and letstrust.de

00:25 <smaeul> have you tried playing with it in Linux with spidev?

00:29 <bauen1> smaeul: because the board is flashed to secure boot, it's not exactly trivial to boot linux and the few times i tried i failed to load the kernel / initrd / fdt correctly, but i'll try exactly that again tomorrow

00:32 <bauen1> and really, i should setup buildroot or something because most kernels e.g. debian don't build the sunxi spi driver, or maybe there is some gpio bit bang implementation in the kernel i haven't found yet, i definietly need to look into it

00:32 <bauen1> somebody else supposedly made soft-spi work on an rpi4 (https://github.com/joholl/rpi4-uboot-tpm)

00:33 <smaeul> yes, there's spi-gpio in Linux

00:38 <apritzel> smaeul: so the RTC gets stuck when trying to set it, and doesn't make progress, which looks like the RTC input clock is not ticking

00:40 <apritzel> when I do hwlock -w, it times out with "Failed to set rtc time.", after writing SUN6I_LOSC_CTRL_RTC_HMS_ACC

00:40 <smaeul> what is in clk_summary?

00:41 <apritzel> then I see the h:m:s part being set, but the date is still the old one. And the time does not move

00:42 <apritzel> bit 8 in reg 0x00 is set, which is why it times out

00:42 <apritzel> clk_summary looked sane on the first glance, I saw the ccu-sun6i-rtc driver clocks there

00:43 <apritzel> but I was wondering that the actual RTC clock is not really referenced anywhere

00:45 <apritzel> smaeul: the RTC driver patch is really simple, literally: https://github.com/apritzel/linux/commit/6dbe675b8be

00:47 <smaeul> Maxime suggested leaving the RTC-internal clocks out of the clock provider, because they aren't useful to anything else

00:47 <apritzel> the DT part is this: https://github.com/apritzel/linux/commit/c665f1230c7b, line 492

00:47 <apritzel> makes some sense, because they would need to be self-referenced inside the RTC node, I guess

00:48 <smaeul> and because the RTC driver can just writel directly without going through the common clock framework

00:48 <apritzel> As far as I can tell the bus block is fine, because I could access the RTC registers just fine via /dev/mem

00:48 <smaeul> but what about iosc and osc32k?

00:49 <smaeul> I'm interested in clk_summary so I could see those enable counts

00:49 <apritzel> yeah, need to fire that TV box up ...

00:50 <apritzel> (my OPi Zero2 seems busted, so it's all a bit more tedious)

00:50 <smaeul> ah

00:52 <smaeul> in any case I think we need a consumer (or a flag) for CLK_RTC_32K so it doesn't get disabled, but I don't know where to put that, or what parent it should have

00:54 <apritzel> right, I was thinking in the same direction

00:58 <apritzel> smaeul: https://pastebin.com/AaPV1zi0

00:59 <apritzel> hwclock is stuck at "Fri Jan 2 00:01:36 1970 0.000000 seconds", so I guess it was running, but then Linux turned it off

01:00 <apritzel> but the RTC driver has no shortcut into the clock driver anymore, right? So we just can't reach into that and turn it on

01:04 <apritzel> yeah, clk_ignore_unused fixes it

01:06 <apritzel> the difference is clk_summary seems to be the very first clock: "iosc": hardware enable is "Y" with clk_ignore_unused, but "N" without it

02:03 <bauen1> i may have been able to "request the 0 locality" from the tpm by writing the a bit offset by one, and then release it again, by doing the same, and the bit indicating that the locality was "active" was also offset by one, this is a timing issue isn't it :(

02:53 <smaeul> apritzel: thanks for confirming this. I think I will put CLK_IS_CRITICAL on CLK_RTC_32K, since the RTC should always be running.

11:02 <NekoMay> dl.linux-sunxi.org has a (very) expired SSL certificate

11:07 <NekoMay> https://mega.nz/folder/lx4CyZBA#PiFhY7oSVQ3gp2ZZ_AnwYA

11:09 <NekoMay> Oops, wrong channel, sorry

11:09 <NekoMay> That's for Sipeed's Lichee RV stuff

11:10 <NekoMay> Meant to send it to someone else

14:26 <swiftgeek> how do I access PL11 on mainline linux?

14:27 <swiftgeek> err nvm

14:27 <swiftgeek> typed in 364 which doesn't exist

14:27 <swiftgeek> 363 works fine

17:39 <ignne> Hi, does anyone here run an A20 and can tell me what current it usually idles at? I'm also interested in standby (s2idle) and everything power-saving related. Thanks!

17:39 <ignne> I had alread asked this a couple of days ago, maybe someone now knows something :)